Featured Projects
Satellite Image Super Resolution
get in touch
2.5 M
Target Resolution
Downscaled from 10m Sentinel-2 images.
> 2 MIO $
Savings
Cost for VHR images for a single take of a mid-sized country.
High-resolution satellite imagery is vital for applications like land-use mapping, environmental monitoring, and disaster response. But the cost is often prohibitive—archival imagery for a mid-sized country can run into the millions of dollars, making many projects infeasible.
Super-resolution offers a cost-effective alternative. By leveraging deep learning, we upscale free low-resolution satellite images to near-commercial quality. This isn’t magic—it’s using deep-learning models and multiple overlapping low-resolution images to reconstruct high-detail images.
For this project, the Asian Development Bank, is developing an open-source super-resolution model and image database that has been rigorously tested in real-world applications. This innovation enables organizations to access high-quality satellite data without the high price tag.
Modelling a bike-sharing scheme
get in touch
12,000+
Number of bikes
One of the largest bike-sharing schemes in the world
800+
Number of docking stations
Covering all of central London
Bike-sharing is a great addition to the urban mobility mix – providing an affordable, sustainable, and flexible transport option. But keeping thousands of bikes in the right place at the right time, without excessive operational costs, is a major challenge.
When a transport company bid for the re-tendering of the London bike-sharing scheme (12,000 bikes), they needed to ensure operational efficiency while meeting Transport for London’s strict service level agreements.
This project involved building a detailed simulation model to predict how bikes flow between stations, identifying when and where interventions were needed. Our insights helped the company refine their operational strategy and optimize their bid.